A Novel Approach to Multi-Document Summarization with Sentiment Analysis

This project presents a new method for automatically summarising a collection of documents with semantic labels. The proposed method uses document clustering to group similar papers together and then generates summaries by combining information from the documents in each cluster. The resulting summaries are then used to generate sentiment classification and obtain a semantic mean for each cluster, along with top negative and positive semantic labels. The proposed method is based on a hybrid summarization model that uses deep neural networks to learn complex relationships among the documents and to fuse disparate features commonly used in multi-document summarization tasks. By introducing a new dataset and addressing challenges in multi-document summarization, this project aims to improve the ability to extract relevant information with sentiment from large volumes of text data.